We Broke Bing's AI Image Generator...

23 Mar 202319:22

TLDRIn this video, the hosts explore Bing's new AI image generator, powered by OpenAI's DALL-E 2. They discuss the integration of AI in Microsoft's search engine, showcase various image generation prompts, and highlight its strengths and limitations. They test the generator with different prompts, from a dog flying an airplane to a human hand holding a knife. The hosts also touch on the content policy that blocks certain political figures and high-profile names. The video concludes with reflections on the future of AI image generation and its potential impact.


  • 😀 Microsoft has integrated an AI image generator named Dali 2 into Bing, following the success of their AI chat functionality.
  • 🤖 The new Bing, with its AI chat feature, has garnered significant attention for Microsoft's search engine, something that hasn't happened in a long time.
  • 🖼️ Users can generate images by providing textual descriptions, such as 'a dog flying an airplane', to the AI image generator.
  • 💡 The chat feature within the image generator can suggest ideas or provide feedback on the prompts given by users.
  • 🚀 The AI can create images quickly, sometimes within seconds, but the quality and accuracy can vary based on the complexity of the request.
  • 💰 There is a limit to the number of image generations a user can make without additional payment, referred to as 'Boost'.
  • 🛑 Certain prompts are blocked by Bing's content policy, such as generating images with political figures in specific contexts.
  • 🔮 The AI's ability to generate images can be unpredictable, with some terms leading to 'broken' or unexpected results.
  • 🎭 The AI seems to handle abstract concepts better than detailed, specific requests, which can lead to more successful image generation.
  • 👀 There is speculation about the transparency and updating of the database that the AI uses to generate images, and how it determines what is allowed or not.
  • 🚫 The system has a content warning and blocking mechanism for sensitive or controversial figures, as seen with attempts to generate images of certain politicians.

Q & A

  • What is the main topic discussed in the video script?

    -The main topic discussed in the video script is Microsoft's Bing search engine's new AI image generator feature and its capabilities.

  • What is the significance of the AI image generator in Bing?

    -The AI image generator in Bing is significant as it represents a new integration with other AI technologies and could potentially make Bing a more advanced search engine, attracting more users.

  • What is the role of 'Dolly' in the context of the video script?

    -In the context of the video script, 'Dolly' refers to an AI image generator that has been integrated into Bing, following the integration of other AI technologies like Chat-GPT and DALL-E.

  • How does the AI image generator work according to the script?

    -The AI image generator works by taking textual prompts and generating images based on those prompts. Users can input descriptions and the AI will create images that match the description.

  • What is the 'Boost' feature mentioned in the script?

    -The 'Boost' feature mentioned in the script is a function within the AI image generator that presumably improves the quality or speed of the image generation process, but it requires a certain amount of 'lightning bolts' which may be a form of in-app currency or points.

  • What are some of the limitations or restrictions the script mentions regarding the AI image generator?

    -The script mentions that certain prompts may lead to content policy violations, which could result in automatic suspension of access. It also notes that some specific terms or names related to politics or controversial figures may be blocked or flagged by the system.

  • What is the implication of the script mentioning a user getting 'banned' for certain image prompts?

    -The implication is that the AI image generator has content moderation in place to prevent the generation of inappropriate or sensitive images, and users who repeatedly violate these policies may face consequences such as being banned from using the service.

  • What does the script suggest about the AI's ability to understand complex or abstract prompts?

    -The script suggests that the AI has varying levels of success with complex or abstract prompts. Some prompts result in quick and seemingly accurate image generation, while others are more difficult for the AI to interpret, leading to less accurate or bizarre results.

  • How does the script explore the AI's response to different types of prompts?

    -The script explores the AI's response by giving various examples of prompts, ranging from simple to complex and abstract, and observing the AI's generated images. It also notes the AI's reaction to prompts that may be considered sensitive or controversial.

  • What is the conclusion the script draws about the AI image generator's capabilities and potential?

    -The script concludes that the AI image generator is a cool and interesting feature, but it also has its limitations and is still in the process of improvement. It also suggests that the AI's understanding and generation capabilities can be unpredictable and are subject to content moderation.



🤖 Introduction to AI Image Generator in Bing

The script introduces a new AI image generator feature integrated into Bing, Microsoft's search engine. It discusses the success of Bing's chatbot with its AI chat functionality and the potential of combining this with Dali, another AI product. The presenter suggests using the image generator to create unique images by providing prompts and discusses the limitations of the system, such as the need to purchase 'Boosts' to generate more complex images. The conversation also touches on the integration of Unreal Engine, and the presenter provides a live demonstration of generating images with various prompts, highlighting the quick response time and the challenges of creating more complex images.


🚫 Content Policy and AI Image Generation Limitations

This paragraph delves into the content policy restrictions of the AI image generator, showing how certain prompts related to violence or sensitive political figures can lead to content warnings or outright bans. The script discusses the predictability of these limitations and the potential for improvement. It also explores the boundaries by testing the system with various prompts involving politicians and controversial figures, revealing that certain names and scenarios are blocked due to content policy violations. The conversation raises questions about transparency, database updates, and the parameters set by the AI system.


🧐 Exploring AI's Perception of Public Figures

The script continues to experiment with the AI image generator, focusing on how it perceives and generates images of public figures, including politicians and influential individuals. It discusses the AI's ability to create images of historical figures and tests its limits by inputting names of living politicians and other notable personalities. The conversation reveals that certain names are flagged and blocked, suggesting a level of censorship or sensitivity filtering in the AI's database. The presenter also speculates on the potential reasons behind these restrictions and the implications for AI-generated content in the future.


🛑 Final Thoughts on AI Image Generation and Content Policy

In the final paragraph, the script wraps up the discussion on AI image generation with a focus on the content policy and its impact on creativity and expression. It reflects on the experiment conducted throughout the video, highlighting the AI's limitations and the potential consequences of pushing the boundaries. The conversation touches on the ethical considerations of AI content generation, the importance of understanding the parameters set by the system, and the need for transparency in how these systems operate. The script concludes with a contemplative note on the future of AI and its role in shaping our digital experiences.



💡AI Image Generator

An AI image generator refers to a technology that uses artificial intelligence to create images based on textual descriptions. In the context of the video, Bing's AI image generator is a feature that allows users to input text prompts and receive generated images. The script discusses the integration of this technology into Bing's search engine, showcasing its capabilities and limitations.


Microsoft is a leading technology company known for its software products and services. The video mentions Microsoft's involvement with AI, specifically in the development of Bing's AI image generator. The script highlights how this feature has brought attention to Microsoft's search engine, which is a significant development in the company's ongoing efforts to innovate and compete in the tech industry.


DALL-E is an AI model developed by OpenAI that is capable of creating images from textual descriptions. The script refers to 'Dali' as a part of the AI technology stack that Microsoft is integrating into Bing, suggesting that similar technology to DALL-E is being used for Bing's image generation capabilities.

💡Chat GPD

Chat GPD, or Chatbot Generative Pre-trained Transformer, is a type of AI-powered chatbot that can generate human-like text responses. The script mentions that Bing has integrated a chat function powered by AI, which is likely a reference to Chat GPD technology. This integration is part of Bing's efforts to enhance user interaction and provide more dynamic search experiences.


In the context of the video, 'Boost' refers to a feature within Bing's AI image generator that allows users to prioritize or speed up the image generation process, possibly at an additional cost. The script discusses the use of Boost to expedite the creation of images and the implications of running out of Boost, which suggests a potential monetization strategy for the service.

💡Content Policy

Content policy refers to the guidelines or rules that govern the type of content that can be created or shared on a platform. The script highlights instances where certain prompts, such as politically sensitive figures, are blocked by Bing's AI image generator due to conflicts with the platform's content policy, demonstrating the limitations and restrictions placed on AI-generated content.

💡Unreal Engine

Unreal Engine is a game engine developed by Epic Games, known for its high-quality graphics capabilities. The script briefly mentions Unreal Engine, possibly as a comparison to the quality of images generated by Bing's AI, suggesting that the generated images are of a high standard that could be likened to those produced by professional game engines.

💡Mid Journey

Mid Journey is likely a reference to the company or technology that powers the AI image generation in the video. The script discusses how certain terms or prompts can lead to predictable failures in image generation, indicating that the technology, while impressive, still has limitations and is in a 'mid-journey' of development and improvement.

💡Sith Lord

A Sith Lord is a character archetype from the Star Wars franchise, known for their dark side powers and villainous roles. The script uses 'Sith Lord' as a creative prompt for the AI image generator, exploring the AI's ability to interpret and visualize abstract and fictional concepts, and noting the AI's responses to such prompts as part of the video's exploration of the technology's capabilities.

💡Elon Musk

Elon Musk is an entrepreneur and CEO known for his work with companies like Tesla and SpaceX. The script mentions trying to generate an image of 'Elon Musk riding a bike' as a prompt for Bing's AI, which is blocked, indicating the AI's content policy restrictions on certain individuals, and sparking a discussion on the parameters and limitations of AI-generated content.


Bing now features an AI image generator, showcasing Microsoft's integration with AI technology.

The new Bing with chat functionality powered by AI has been a significant success, sparking renewed interest in Microsoft's search engine.

Dolly, Microsoft's image generator, is being integrated into Bing, suggesting a potential future for the search engine.

The integration of Dolly with Bing's AI chat functionality is part of a broader AI strategy by Microsoft.

The AI image generator allows users to create images with text prompts, such as 'a dog flying an airplane'.

Some generated images may not fully align with the prompt, indicating the AI's limitations in understanding complex instructions.

The AI struggles with generating images that involve violence or sensitive political figures.

The AI image generator uses a 'Boost' feature to enhance image generation, which may come at a cost to users.

The AI's response time varies depending on the complexity of the image prompt, with some taking longer to generate.

There are content policy restrictions in place that prevent the AI from generating images of certain political figures.

The AI can generate abstract images more successfully than those with detailed or realistic expectations.

Users need to sign in to use the AI image generator, indicating some level of user engagement tracking.

The AI's ability to generate images of non-politicians and fictional characters is less restricted.

The AI image generator's response to prompts involving controversial figures or sensitive topics is blocked or flagged.

There is speculation about the transparency and updating of the AI's content policy database.

The AI's generated images can sometimes be unpredictable, even with seemingly simple prompts.

The AI image generator offers a glimpse into the future of AI and its potential impact on content creation.